This conference paper was submitted for presentation at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

In the rapidly evolving landscape of engineering certification, traditional methodologies are becoming increasingly inadequate as industries advance towards more innovative and high-performance designs. Existing certification workflows in fields such as automotive and aerospace rely heavily on factors-of-safety estimations and practices derived from historical data and physical testing, which can be costly and time-consuming [1] . Derivative methods, outlined in standards like NASA-STD-5001B [2], depend on archives of legacy data, which may not adequately address novel materials and designs emerging from advancements in structures and manufacturing techniques. Moreover, while test-based certification in standards like ASTM D3039/D3039M-17 [3] provides more accurate predictions of material behaviour through prototype testing, it is limited under specific environmental conditions and becomes inadequate for complex system interactions and greater variability in operating environments [4]. These methods prescribe conservative approaches, including factors of safety and prototype testing, often resulting in unnecessarily heavy structures that do not optimize performance or guarantee safety [4]. Digital twin technology offers a powerful solution to traditional certification challenges by creating virtual representations of physical assets that adapt using real-time data. Our study introduces a digital twin model developed by Autodesk Research that uses strain gauge signals to generate dynamic stress outputs across an entire system, even with unknown stimuli. By integrating sensors into a composite load-bearing structural member of an Unmanned Aerial Vehicle (UAV), a test rig is created to mimic in-lab conditions thorough testing. This approach allows for real-time stress field visualization through principled physics-based data extrapolation from strategically placed sensors. Our study consists of three components. The first component employs Autodesk Nastran as a Finite Element Analysis (FEA) solver, allowing for the development of a fully functional digital twin without the need for an in-house FEA solver. The second component involves a three-way validation between Nastran, the digital twin algorithm, and physical sensors. By applying controlled loads to the test rig, we can replicate the conditions in both Nastran and the digital twin, then compare resulting stress fields against physical sensor data collected from the test rig. In the third component, the comparison between Nastran, the digital twin, and sensor data are repeated without assuming knowledge of the load. Instead, the digital twin algorithm is augmented with the ability to infer the operating conditions from in-situ sensor data. This approach helps close the correlation gap between sensor data, simulation, and the digital twin, informing critical design decisions by providing insights into the accuracy of assumptions and predictions. Our goal is to integrate the structural member from the test rig into the UAV system using modal reduction techniques. This allows for visualization and simulation of the structural member using a robust, physics informed digital twin accurately replicating in-flight conditions. By repeating this process for all components of the UAV, we aim to create a fully developed sensorized digital twin, enabling collection of real-time dynamic data and assessment of structural responses during operation. Unlike conventional testing, which often relies on assumed maximum loads with uncertain accuracy, this method reduces uncertainty by delivering estimated loads and accurate structural responses. This clarity enhances the understanding of safety margins and addresses current challenges in design and certification. The monitoring capabilities of this digital twin helps streamline the certification process, potentially eliminating traditional cyclic testing and iterative evaluations. As confidence in system behavior grows, this approach opens innovative design possibilities for composites and other materials, moving beyond reliance on archived data and physical testing while improving simulation accuracy. This methodology establishes operation-specific conditions tailored to unique use cases, paving the way for advanced designs in highly sensitive applications.